As I posted in yesterday’s blog post, I am trying to figure out why CoreNLP doesn’t like being run as a service on Knuth. I’m starting off by isolating the problem: is it the CoreNLP server itself that doesn’t like the multiple requests? Or is it the CoreNLP libraries themselves? I had a hunch that it’s the former, and I confirmed that by creating a separate script that runs the exact same thing as the server does. So, at this point, I am thinking that it has to do with the way that the Stanford CoreNLP server code handles the thread pool of annotators. Tomorrow, I am going to download the entire CoreNLP source code to Knuth and try to debug more with print statements. Really hope to have an answer to this by the end of tomorrow. Meanwhile, I am still waiting for that quota bump on Knuth.